Could my 4790k be causing my random overheating problems?

[McGraw]

Hi.

I've got a very odd and frustrating problem and before I RMA my board I'm keen to see whether there's even the slightest chance it could be my CPU at fault. I can't let my family use this living room pc and I can't start using it without checking temps are ok.

The board is a Gigabyte Z97n Wi-Fi.

Randomly, probably between every 5 and 10 shutdowns (not restarts or sleep resumes) on average, my CPU starts creeping towards 100c in bios and about 60c in idle in Windows at 800mhz. If you try and do anything demanding, the CPU starts throttling almost instantly. The temps are real, I can feel from the coolers I've tried (H60 and stock intel).

I have narrowed it down to the board or CPU and I can't test either myself. I've tried different ram, psu, heatsinks, with IGP, with discrete GPU and it happens without a hard drive or any unecessary connections at all such as USB or audio and mounted on a cardboard box.

I've also run at stock, overclocked, all BIOS versions, optimised defaults, new battery and flashed main BIOS with backup. It's definitely the board or CPU and I have checked and there is 100% no damage to either from cooler installation (I have photographed both and studied in detail).

If the problem doesn't occur at boot, it never appears during that session. Fortunately this means the pc is usable once I'm sure all is well.

Just to throw a spanner in the works, I've found out by pushing the power button in panic at the throttling, that a sleep cycle of Windows cures the problem until the next random time it occurs. If you don't sleep cycle Windows, the problem can only be got rid of in bios by messing with the bios (not sure exactly what solves it as it seems different each time). You can change settings in BIOS then put them back and save and it can cure it. If you continue to Windows the problem remains if it wasn't sorted in BIOS. Obviously it's much easier and more effective to let Windows boot then sleep/wake cycle it.

I have studied bios readings and HWmon/AIDA/Core temp/Speed fan readings and I cannot see any reason whatsoever for the CPU to be overheating. Voltages and fan speeds are identical with or without the problem.

It's almost like the motherboard sometimes supplies far more volts than it thinks it's doing.

I wish I knew what happened exactly when a normal Haswell sleep cycle takes place, it would certainly help me to decide.

I also can't get rid of the nagging feeling that somehow Windows is corrupting my UEFI and a could help...for now though, let's just consider CPU and mobo and assume that hasn't worked (but I will try it).

So, you've got a fistful of cash to bet on the cause, do you bet on CPU or motherboard?

Thanks a lot.

 

[ShintaiDK]

Are you sure the cooler is seated correctly and the fan running?

But it shouldnt restart due to heat throttle.

Even with the stock cooler, the CPU should idle around 40C or so with quiet mode fan settings.

 

[McGraw]

Thanks but I'm 100% that the cooler(s) are not the issue. A sleep/wake cycle of windows instantly puts temps back to where they should be.

 

[BSim500]

Randomly, probably between every 5 and 10 shutdowns (not restarts or sleep resumes) on average, my CPU starts creeping towards 100c in bios and about 60c in idle in Windows at 800mhz.

60c idle is way off. Should be nearer 30c. Either your heatsink isn't seated correctly, or something's wrong with the CPU. Does the heatsink itself ever feel warm to the touch under load? If it were just a case of a faulty temp sensor, then your CPU wouldn't shutdown.

Edit : If there's no way to test one vs other (perhaps you know someone with a Haswell you could test temporarily - doesn't have to be an i7 could be an i5/i3/Pentium - just to check if the board doesn't go screwy with that), and you bought them recently then personally, I'd RMA both (simply because some will charge a restocking fee for a perfectly working item if you send the "wrong" one in, plus it will give them the chance to see what's wrong).

Edit 2 : Exactly what is the vcore being fed to it when idle? Have you tried to force a lower "fixed" vcore in the BIOS? A faulty motherboard might be misreading the CPU's VID table?

 

[Idontcare]

Thanks but I'm 100% that the cooler(s) are not the issue. A sleep/wake cycle of windows instantly puts temps back to where they should be.

Unfortunately you can't rule out a mis-seated cooler until you've physically inspected the backside of the mobo and visually observe that all four retention pins are fully in place.

Short of that, anything you would do to convince yourself the cooler is OK (such as the sleep/wake cycle you envision) is simply going to be a crutch that wastes your time by fooling you into thinking you've ruled out the cooler.

There is a reason you will see many many posts in your thread telling you your cooler is probably not mounted correctly, it is because we've all been down this road before. We've all convinced ourselves 100% the cooler could not be the problem, then we wasted a bunch of our own time (plus that of the people we asked to help us) trying to debug other parts of our rig or software, only to eventually come to figure out that the cooler simply isn't mounted correctly.

 

[BonzaiDuck]

Unfortunately you can't rule out a mis-seated cooler until you've physically inspected the backside of the mobo and visually observe that all four retention pins are fully in place.

Short of that, anything you would do to convince yourself the cooler is OK (such as the sleep/wake cycle you envision) is simply going to be a crutch that wastes your time by fooling you into thinking you've ruled out the cooler.

There is a reason you will see many many posts in your thread telling you your cooler is probably not mounted correctly, it is because we've all been down this road before. We've all convinced ourselves 100% the cooler could not be the problem, then we wasted a bunch of our own time (plus that of the people we asked to help us) trying to debug other parts of our rig or software, only to eventually come to figure out that the cooler simply isn't mounted correctly.

I've had better luck than that, but I know pretty well what can happen in cooler installation. The stock coolers can be just a bit difficult, and you have to look carefully at the split arrow-shaped pins to make sure they're secure. There are almost as many ways that people make mistakes with their HSF installations.

But here's a casual thought. Unless he says otherwise, I'm assuming the OP is using the Corsair Hydro H60. So I'll make a couple assumptions, and OP can correct me on one or both. If he used the H60 in addition to the stock cooler, then he's more likely to use a similar AiO cooler if the H60 has been replaced. And if it's the H60, then . . . . It's the H60.

He says that restarting the computer (or variations of power-cycling) temporarily render normal operation, which somehow reverts to the status of a hotter-than-normal CPU. He says that he can feel the heat from the cooler(s).

People had posted in past years complaining that the pump on their AiO had failed. I thought maybe one of those cases was an H100. So I wonder if it isn't possible that a pump might show partial failure, some electrical connection that shows improvement on a complete power-cycle, or some such related situation.

It's easy to have wishful thinking about an item of installation inconvenience.

On the other hand, I've begun to notice at least a few posts with Haswell thermal anomalies. And -- Haswell-E.

 

[McGraw]

It's not my cooler.

I will say it again...my temps are fine for days on end then suddenly they won't be after a cold boot.

When I sleep my PC and wake it again the temps return to normal...I will film this behaviour next time if I'm still not being believed.

They then stay normal for a few more days then go crazy again after 5 cold boots or so.

This happens with the stock cooler too.

I've had the PC built outside its case to check a proper seat and it's absolutely fine.

Volts read what thy should whether stock, clocked, idle or load.

I don't want to be persuaded that it's my cooler I just want an idea of whether my cpu could be faulty.

Ps it's not my cooler or the seating of said cooler.

I don't know anyone who has a haswell set up and I don't know whether I would want to risk their kit anyway.

I'm going to rma the board unless someone can give a good argument as to how it could be the CPU.

 

[mistersprinkles]

Sounds like you have a defective board or a defective CPU. My money is on defective board. Defective CPU is extremely rare. I'd RMA both just to be safe.

 

[McGraw]

Ok I've just noticed the post above my last one and that is definitely the most helpful reply...however the stock cooler does the same so it can't be a case of the pump working properly after a power cycle.

Perhaps one of the fan headers or the connection to the motherboard is faulty.

I just can't see it being the CPU but I want to make sure as I the board from a different place to the CPU.

Ps I'm using the h60 and right now my temps are great.

 

[McGraw]

Sounds like you have a defective board or a defective CPU. My money is on defective board. Defective CPU is extremely rare. I'd RMA both just to be safe.

Thanks. It has to be one of them but I really want it to be the motherboard as the retailer is much easier to deal with.

 

[BonzaiDuck]

It's not my cooler.

I will say it again...my temps are fine for days on end then suddenly they won't be after a cold boot.

When I sleep my PC and wake it again the temps return to normal...I will film this behaviour next time if I'm still not being believed.

They then stay normal for a few more days then go crazy again after 5 cold boots or so.

This happens with the stock cooler too.

I've had the PC built outside its case to check a proper seat and it's absolutely fine.

Volts read what thy should whether stock, clocked, idle or load.

I don't want to be persuaded that it's my cooler I just want an idea of whether my cpu could be faulty.

Ps it's not my cooler or the seating of said cooler.

I don't know anyone who has a haswell set up and I don't know whether I would want to risk their kit anyway.

I'm going to rma the board unless someone can give a good argument as to how it could be the CPU.

I don't dismiss these other possibilities. In fact, you can check recent threads here, in "mobos" and "Cases/Cooling." There are a lot of "question marks" over cooling symptoms and temperatures; there's mention or rumors of X99 boards "smoking and catching on fire." These are new chipsets that share a lot of features -- X99 and Z97. Also with the processors, except the 4790K has an "improved polymer thermal interface material" instead of indium solder.

Wish I could offer more help. If you have no way to test the board and CPU by replacing one or the other temporarily, I suppose you could RMA both items and start over with your project. It could only be a couple months old, right?

 

[crashtech]

Accidental overvoltage is the only thing I can think of that would cause these symptoms, which would appear any time the CPU is NOT throttled down by EIST.

It would be interesting to manually lock the voltage in BIOS to 1.15-1.2V or so and see what happens.

 

[BonzaiDuck]

Accidental overvoltage is the only thing I can think of that would cause these symptoms, which would appear any time the CPU is NOT throttled down by EIST.

It would be interesting to manually lock the voltage in BIOS to 1.15-1.2V or so and see what happens.

"Accidental?" That's possible with the Haswell-gen boards? My point of reference is 3+ years old. . .

Someone else pointed to the relative probabilities -- board vs CPU. I've been too suspicious of Intel of late, and it is certainly true that it is most unlikely you would ever get a defective processor. Not impossible, but most unlikely.. .

 

[Termie]

Ok I've just noticed the post above my last one and that is definitely the most helpful reply...however the stock cooler does the same so it can't be a case of the pump working properly after a power cycle.

Perhaps one of the fan headers or the connection to the motherboard is faulty.

I just can't see it being the CPU but I want to make sure as I the board from a different place to the CPU.

Ps I'm using the h60 and right now my temps are great.

Sounds like a bad pump, actually. That would cause overheating immediately, only to be reset with a sleep/wake cycle.

Can you confirm that you've actually had the EXACT same behavior with the stock Intel heatsink?

If you can confirm that you saw this (as it's not clear from your first post), then my bet is that the motherboard's voltage regulation has gone haywire.

 

[BonzaiDuck]

Sounds like a bad pump, actually. That would cause overheating immediately, only to be reset with a sleep/wake cycle.

Can you confirm that you've actually had the EXACT same behavior with the stock Intel heatsink?

If you can confirm that you saw this (as it's not clear from your first post), then my bet is that the motherboard's voltage regulation has gone haywire.

With all the other stuff popping up in Haswell threads in these forums, I keep asking myself why Intel would decide to jettison FIVR with Skylake.

I suppose the OP will have to look more carefully at the system working with both the two coolers.

[ . . . Despite what others have told me today on these forums, I wonder about the voltage regulation on these cores. . . ]

 

[McGraw]

Thank you for the further comments.

Although I have definitely seen this behaviour with the stock cooler I have decided that just for the record I am going to strip it all down again and have one last try seeing as the board will be going back in the box anyway.

One thing I may have not made clear is that behaviour never occurs from a restart, only from shutdown.

If I can get a confident approximation of how many shutdown/power off cycles this can be replicated in, I will rma the board and ask them to see what they can find.

If they don't find a problem I will rma the cpu.

If the cpu supplier doesn't find a problem then I'm stuck!

When this first started happening it wouldn't be resolved by a reboot but now it sometimes will so it's going to be a hassle replicating it.

 

[DrMrLordX]

Do you have any tools to measure power draw from the wall, such as a Kill-a-Watt? I know Kill-a-Watts aren't the most accurate things in the world (particularly when it comes to power spikes), but any sustained increase in power draw should register on one.

Such a device should be able to show you differences in power draw between "normal" operation (idle, for example) and overheat conditions. If you see a difference, then you know some part of the system is sucking up more juice than "normal".

 

[tolis626]

I know I'm probably late, but I thought I'd throw this out there. Are you using any software in Windows that changes voltages/clocks? I don't know if Gigabyte has such a thing, but there have been instances when AISuite on my rig will go nuts and max a slider out, be it iGPU or CPU voltage or maybe even uncore voltage. So yeah, it might be stupid, but check on the software side, just to be sure. And when I say software side, also check the BIOS. Maybe an "easy overclocking" feature or something is going wrong.

I'd do what others said and try with locked voltage. If nothing helps, I'd put my money on the board too. Even if your CPU was somehow defective, I really doubt that we'd see such symptoms.

 

[McGraw]

Thanks.

The problem occurs in bios and continues to windows so I know it can't be software. Plus I only overclock in windows.

The killawatt thing is a good idea too. Will pinch my dad's when I see him next.

 

[crashtech]

There are many things that could be done and that I would do before swapping parts, but I think the OP just wants to replace something. So replace the motherboard, that is the most likely source of the problem.

 

[BonzaiDuck]

Thanks.

. . . Plus I only overclock in windows.
. . . . .

The prevailing veteran wisdom here in the forum will shift and change with the technology. I'm running an ASUS board, and the general thoughts about overclocking seem to follow through the latest chipsets. [some may correct me if I'm wrong.] And a lot of ASUS users seem to ritually throw verbal poop at the software AI Suite. The Suite has many virtues, but I've yet to hear anyone in the forums tout the practice of overclocking through Windows software as opposed to hands-on changes to the BIOS itself.

Also -- some boards (like mine) keep track of a "sequence" of overclock adjustments, counting the settings with the most recently used method as the operative settings. I could overclock through the BIOS, then use the Windows software -- which would then give the operative settings.

I just advise to use BIOS directly.

 

[crashtech]

I can't shake the impression that the voltage has been set to something outrageous and it being misreported in the OS, so then the only thing saving it is power management, which is why it overheats when in the BIOS.

 

[cytg111]

My 4770(-non K) will throttle at heavy avx2 loads running the stock cooler. IIRC it idles around mid to high 30'ies .. If you are idling at 60 degrees celsius, something is really really wrong, our your ambient temps is like 50..

 

[BonzaiDuck]

My 4770(-non K) will throttle at heavy avx2 loads running the stock cooler. IIRC it idles around mid to high 30'ies .. If you are idling at 60 degrees celsius, something is really really wrong, our your ambient temps is like 50..

This had also raised the issue about which stress-tests to use. Some folks are validating their overclock with limp stability tests because of the temperatures. I have no firsthand experience on the Haswell cores and motherboards z87, z97 or x99.

It just seems -- in hindsight -- that we had the "TIM" issue beginning with IB, and "fully integrated voltage regulation" in Haswell.

Per FIVR, it will get kicked to the curb with SkyLake. Or -- so I heard. I'm not that "electronics savvy." But if there had been a common flaw with old C2D Wolfdales that had the temperature sensor reading stuck at a value way above expected stock idle, how would it be impossible that there is some random flaw in FIVR? Sooner or later, we'll likely know more . . .

 

[PhIlLy ChEeSe]

It's not my cooler.

I will say it again...my temps are fine for days on end then suddenly they won't be after a cold boot.

When I sleep my PC and wake it again the temps return to normal...I will film this behaviour next time if I'm still not being believed.

They then stay normal for a few more days then go crazy again after 5 cold boots or so.

This happens with the stock cooler too.

I've had the PC built outside its case to check a proper seat and it's absolutely fine.

Volts read what thy should whether stock, clocked, idle or load.

I don't want to be persuaded that it's my cooler I just want an idea of whether my cpu could be faulty.

Ps it's not my cooler or the seating of said cooler.

I don't know anyone who has a haswell set up and I don't know whether I would want to risk their kit anyway.

I'm going to rma the board unless someone can give a good argument as to how it could be the CPU.

Why don't you start again, List your full system spec's were not mind readers. Also you come in here asking for help then yell at people who reply? You'll not get much useful help till you do list your full system, along with a I don't know it all attitude! Clearly the cause of the issue is you, RMA is not a fix...............